> For the complete documentation index, see [llms.txt](https://docs.corpdev.ai/corpdev.ai-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.corpdev.ai/corpdev.ai-docs/documents-generation-and-exports-api.md).

# Documents, generation and exports API

All paths below are relative to `https://api.corpdev.ai/v1/public`. Authentication and workspace permissions apply to every protected operation.

## Read and create

| Method and path              | Purpose                                              |
| ---------------------------- | ---------------------------------------------------- |
| `GET /documents`             | List accessible deliverables with cursor pagination. |
| `GET /documents/:documentId` | Read a deliverable.                                  |
| `GET /folders`               | List deliverable folders.                            |
| `POST /documents`            | Create a draft or start a generated deliverable.     |

Creation requires `name`. Optional fields include `markdown`, `docType`, `folderId`, `prompt`, `model`, `contextDocuments` and `cloudFiles`. A non-null folder must exist in the key's workspace. Use references to accessible source documents and files; do not pass another workspace's path as a reference.

Unknown creation fields are rejected. Omit `docType` to use the default document type. A request without a non-empty prompt creates a draft; a prompt starts metered Analyst work. See the [quick start](https://docs.corpdev.ai/corpdev.ai-docs/rest-quick-start) for a complete draft example.

## Continue an Analyst task

`POST /documents/:documentId/prompt` requires a non-empty `prompt` and write access. For example, the JSON body can be:

```json
{
  "prompt": "Add an open diligence questions section. Distinguish missing evidence from confirmed findings."
}
```

Optional context and model options must match the supported schema. Do not assume a second prompt always starts immediately: the response can report queuing or a conflict with current work.

Poll `GET /jobs/:jobId` for a known job, or `GET /documents/:documentId/job` for the document's job. Use the returned status and identifiers. The legacy path `/documents/:documentId/jobs/:jobId` is not the current polling route.

## Revise, copy or remove

| Method and path                         | Purpose                                         |
| --------------------------------------- | ----------------------------------------------- |
| `PUT /documents/:documentId`            | Update supported document fields.               |
| `POST /documents/:documentId/apply-ops` | Apply supported structured document operations. |
| `POST /documents/:documentId/duplicate` | Create a separate copy.                         |
| `DELETE /documents/:documentId`         | Delete the document.                            |

For version-sensitive full replacement operations, use `expectedVersion` and handle HTTP 409 rather than overwriting changes blindly. Confirm the operation schema before writing. A delete is not an archive or a reversible draft change.

## Attachments and slides

`GET /documents/:documentId/attachments` lists attachments. Read a particular attachment with `GET /documents/:documentId/attachments/:attachmentId`. Upload attachments using `POST /documents/:documentId/attachments` with multipart field `files` and write access. Let the HTTP client generate the multipart boundary.

`GET /documents/:documentId/slides` reads slide content where present. An attachment or slide result can differ from the main document's Markdown; inspect the intended representation.

## Export

`POST /documents/:documentId/exports/:format` requires `export` scope and returns the exported file. Supported format names are `markdown`, `slides-markdown`, `docx`, `pptx` and `xlsx`. An XLSX export reads the document's stored workbook; if no workbook exists, it returns 422. Do not assume every format offered elsewhere in the product is available on this route.

The default response is the file; requesting `Accept: application/json` returns a JSON representation containing base64 content. Save the file using its content type and filename. For Word exports, inspect `X-Export-Fidelity` when present: a fallback render can have reduced fidelity. Handle a non-success response before treating its body as a document. Open generated Office files to check their layout before distribution.